## Table 4 |
||||||

Performance of various cleavage prediction models to predict cleavage in pig prohormones |
||||||

Performance |
Known |
Mammalian |
Human |
Logistic |
Human |
ANN^{d} |

Criteria^{a} |
Motif |
Logistic |
AA^{b} |
AA Prop.^{c} |
AA |
AA Prop. |

True Positives | 181 | 165 | 160 | 158 | 164 | 167 |

True Negatives | 1520 | 1640 | 1724 | 1670 | 1735 | 1747 |

False Positives | 329 | 209 | 125 | 179 | 114 | 102 |

False Negatives | 54 | 70 | 75 | 77 | 71 | 68 |

Correct Classification | 0.8162 | 0.8661 | 0.904 | 0.8772 | 0.9112 | 0.9184 |

Sensitivity | 0.7702 | 0.7021 | 0.6809 | 0.6723 | 0.6979 | 0.7106 |

Specificity | 0.8221 | 0.887 | 0.9324 | 0.9032 | 0.9383 | 0.9448 |

Positive predictive power | 0.3549 | 0.4412 | 0.5614 | 0.4688 | 0.5899 | 0.6208 |

Negative predictive power | 0.9657 | 0.9591 | 0.9583 | 0.9559 | 0.9607 | 0.9625 |

Correlation | 0.4358 | 0.4856 | 0.5645 | 0.4944 | 0.5919 | 0.6184 |

AUC | 0.8006 | 0.847 | 0.86 | 0.8186 | 0.8589 | 0.8802 |

^{a} Performance criteria. True positives: number of correctly predicted cleaved sites;
True negatives: number of correctly predicted non-cleaved sites; False positives:
number of incorrectly predicted cleaved sites; False negatives: number of incorrectly
predicted non-cleaved sites; Correct classification rate: number of correctly predicted
sites divided by the total number of sites; Sensitivity (one minus false positive
rate): number of true positives divided by the total number of sites cleaved; Specificity
(one minus false negative rate): number of true negatives divided by the total number
of sites not cleaved; Positive predictive power: number of true positives divided
by the total number of sites predicted to be cleaved; Negative predictive power: number
of true negatives divided by the total number of sites predicted to not be cleaved;
Correlation coefficient: Mathewâ€™s correlation coefficient between observed and predicted
cleavage; and AUC: Area under the receiver operator characteristic or ROC curve relating
sensitivity and 1-specificity.

^{b} AA: models trained only on amino acids.

^{c} AA prop: models trained with amino acids combined with the physicochemical properties
of amino acids.

^{d} ANN: artificial neural network approach.

Porter * et al.*

Porter * et al.* *BMC Genomics* 2012 **13**:582 doi:10.1186/1471-2164-13-582